Islet autoantibodies as predictors of type 1 diabetes development, a systematic review with public health implications for Africa Authors: Babatunde Ishola Gabriel Adejumo, Afolabi Adiru Adegboye, Folorunso Lawrence Dada, Margaret Oyarazi Adebowale, Fidelis Ohiremen Oyakhire, Grace Ayobami Fajemidagba, Mariachidera Angela Obiajunwa, Nkechi Augustina Olise, Olufunke Deborah Abolarin, Emmanuel Olawale Ogunmola, Uteno Itanyi Drisu, Nnagbo Nonso Udemezue, Otiogha Udogidi Erowo, Ronke Funmilayo Ajiboye, Kenneth Echezona Ilekuba, Lilian Weremenesongha Agindoton Published: 2026 ABSTRACT AbstractBackgroundEarly identification of individuals at high risk of developing type 1 diabetes (T1D) remains a major public health priority. Islet autoantibodies are widely recognised as key biomarkers of autoimmune beta-cell destruction and disease progression; however, their predictive performance across diverse populations, particularly in African settings, has not been comprehensively synthesised. This study systematically reviews and synthesises evidence on the predictive value of islet autoantibodies for T1D development, with particular attention to emerging evidence from Africa.MethodA systematic review was conducted in accordance with PRISMA 2020 guidelines. Literature searches were performed across PubMed, Scopus, Web of Science, and Embase for studies published between 2000 and 2026. Eligible studies included longitudinal and observational investigations assessing the relationship between islet autoantibody positivity and subsequent T1D development. Data were extracted on study characteristics, autoantibody profiles, and effect measures. Qualitative synthesis was conducted for all eligible studies, while a restricted random-effects quantitative synthesis was undertaken for studies reporting sufficiently comparable outcomes. Risk of bias and certainty of evidence were assessed using established frameworks.ResultsForty-seven studies were included in the qualitative synthesis, including eight studies conducted in African populations or specifically addressing African diabetes phenotypes. A restricted subset of studies contributed to the quantitative synthesis. Across these studies, elevated autoantibody titres and multiple-autoantibody positivity were consistently associated with increased risk of progression to T1D. The restricted quantitative synthesis suggested an approximately threefold higher risk among individuals with elevated or multiple-autoantibody profiles (RR = 3.47, 95% CI: 2.57–4.68), although interpretation should be cautious given the limited number of comparable studies. High-affinity autoantibodies measured using advanced assays demonstrated stronger predictive performance. African studies highlighted substantial heterogeneity in disease phenotype, including lower autoantibody prevalence, later age at diagnosis, preserved endogenous insulin secretion, and evidence of non-autoimmune insulin-deficient diabetes, indicating that predictive models developed in high-income settings may not be directly transferable.ConclusionIslet autoantibodies remain important predictors of T1D development, particularly when multiple antibodies are present. However, substantial regional heterogeneity exists, especially in African populations, where atypical diabetes phenotypes are increasingly recognised. These findings support the clinical utility of autoantibody-based risk stratification while underscoring the need for region-specific validation studies, longitudinal African cohorts, and contextually appropriate screening strategies before widespread implementation.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13098-026-02214-z.Keywords: Type 1 diabetes, Islet autoantibodies, GADA, IA-2A, ZnT8A, Disease prediction, Systematic review, Africa Introduction Type 1 diabetes (T1D) is an autoimmune disease characterised by progressive destruction of pancreatic beta cells and lifelong dependence on exogenous insulin therapy. The global burden of T1D continues to increase, with recent estimates indicating substantial growth in both incidence and prevalence among children, adolescents, and young adults worldwide [24, 25]. Systematic reviews further demonstrate sustained increases in T1D incidence across Europe and several other world regions over recent decades [9, 30]. Early identification of individuals at high risk of progression to clinical disease has therefore become a major public health priority, particularly as advances in disease staging and preventive interventions have increased opportunities for earlier diagnosis and risk stratification. However, important regional disparities persist in disease detection, classification, and surveillance, particularly in low- and middle-income countries where diagnostic infrastructure and longitudinal cohort data remain limited. These challenges are especially evident in Sub-Saharan Africa, where the epidemiology and clinical presentation of diabetes frequently differ from classical descriptions derived from European and North American populations. Emerging evidence suggests that a substantial proportion of insulin-deficient diabetes cases in African populations may not exhibit the typical autoimmune characteristics observed in conventional T1D, including lower prevalence of islet autoantibodies and greater preservation of endogenous insulin secretion [13, 14, 34]. Earlier studies from Tanzania, South Africa, Cameroon, and other African settings have similarly reported heterogeneity in autoantibody prevalence and disease phenotype, raising important questions regarding the applicability of prediction models developed predominantly in high-income countries [2, 18, 26]. Recent commentaries further emphasise that diabetes in Africa represents a unique and underexplored challenge requiring context-specific approaches to diagnosis, classification, and prevention [33]. Collectively, these findings highlight the need to evaluate whether established immunological predictors of T1D perform similarly across diverse African populations. Among the most extensively studied predictors of T1D are islet autoantibodies, including insulin autoantibodies (IAA), glutamic acid decarboxylase antibodies (GADA), insulinoma-associated antigen-2 antibodies (IA-2 A), zinc transporter 8 antibodies (ZnT8A), and classical islet cell antibodies (ICA). These biomarkers reflect ongoing autoimmune activity and frequently appear months or years before clinical disease onset. Longitudinal cohort studies have consistently demonstrated that the risk of progression to T1D increases substantially with the presence of multiple islet autoantibodies, while antibody titres, affinity, and persistence further improve risk stratification [11, 15, 22, 23]. Technological advances in autoantibody detection, including electrochemiluminescence-based assays, have further enhanced predictive performance and improved identification of high-risk individuals [6, 10]. Recent studies also suggest that specific autoantibody profiles may provide clinically useful information regarding disease progression, staging, and therapeutic requirements [8]. Consequently, islet autoantibody assessment has become a central component of contemporary T1D prediction frameworks. Despite these advances, important uncertainties remain regarding the translation of autoantibody-based prediction into population-level public health practice. While several high-income countries have introduced screening initiatives and pre-symptomatic disease staging frameworks, implementation remains limited in most African settings [5, 17]. Differences in genetic architecture, environmental exposures, healthcare access, and disease phenotype may influence both the prevalence and predictive value of autoantibody markers across populations [13, 33]. Furthermore, screening programmes must be evaluated not only according to predictive accuracy but also in relation to feasibility, affordability, ethical considerations, and health system readiness [17, 19]. These considerations are particularly relevant in Africa, where healthcare resources are often constrained, and epidemiological evidence remains comparatively sparse. Previous reviews have examined global T1D epidemiology, autoantibody prevalence, and advances in disease prediction [21, 29, 35]. However, no previous synthesis has comprehensively evaluated the predictive value of islet autoantibodies while explicitly examining the implications of emerging African evidence for risk prediction, screening, and public health policy. This represents an important knowledge gap given the growing recognition of atypical diabetes phenotypes and the limited representation of African populations in longitudinal prediction studies. Therefore, this study systematically reviews and synthesises evidence on the predictive value of islet autoantibodies for the development of T1D and evaluates the implications of current evidence for early detection, risk stratification, and future screening strategies in African populations. Materials and methods Protocol and reporting guideline The design, conduct, and reporting of this systematic review were undertaken in accordance with the PRISMA 2020 (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) statement and its accompanying explanation and elaboration documents [27, 28]. The PRISMA framework was adopted to enhance transparency, reproducibility, and completeness of reporting throughout all stages of the review process, including study identification, screening, eligibility assessment, data extraction, qualitative synthesis, and restricted quantitative synthesis. A predefined review framework was established before study selection to guide the formulation of the research question, eligibility criteria, search strategy, study screening procedures, data extraction, and evidence synthesis. The review was based exclusively on published literature and did not involve primary data collection, human participants, or identifiable patient information; therefore, ethical approval was not required. To evaluate the certainty and interpretability of the evidence, elements of the GRADE (Grading of Recommendations Assessment, Development and Evaluation) framework were incorporated into the assessment of evidence quality and confidence in findings [4, 36]. In addition, methodological principles derived from living systematic reviews were considered to facilitate future updating of the evidence base as new studies emerge in the rapidly evolving field of type 1 diabetes prediction and biomarker research [1]. All stages of the review were conducted using predefined methodological procedures designed to maximise consistency and minimise selection bias. Detailed search strategies, study selection procedures, data extraction methods, risk-of-bias assessments, study-level extraction tables, and supplementary materials are provided to enhance transparency, reproducibility, and critical appraisal of the review process. Eligibility criteria Studies were selected according to predefined eligibility criteria based on the Population, Exposure, Comparator, Outcome, and Study Design (PECO) framework. The population of interest comprised children, adolescents, and adults without clinically diagnosed type 1 diabetes (T1D) at baseline but who were prospectively assessed for progression to T1D during follow-up. Studies involving first-degree relatives of individuals with T1D, genetically at-risk populations, community-based cohorts, and general population screening programmes were eligible for inclusion. This broader eligibility criterion was adopted because a substantial proportion of the existing prediction literature originates from family-enriched cohorts, although such populations may not fully represent the wider population at risk of developing T1D. Given the recognised differences in disease phenotype across populations, studies conducted in African settings were specifically retained and examined in greater detail where available [13, 14, 33]. The exposure of interest was the presence of one or more islet autoantibodies associated with autoimmune beta-cell destruction, including insulin autoantibodies (IAA), islet cell antibodies (ICA), glutamic acid decarboxylase antibodies (GADA), insulinoma-associated antigen-2 antibodies (IA-2 A), and zinc transporter 8 antibodies (ZnT8A). Studies evaluating single-autoantibody positivity, multiple-autoantibody positivity, autoantibody titres, affinity, persistence, or combinations of autoantibody markers were considered eligible [11, 15]. Comparator groups included participants without detectable islet autoantibodies, participants with alternative autoantibody profiles, or groups stratified according to autoantibody number, titre, affinity, or risk category. Studies without a clearly defined comparator or longitudinal risk assessment were excluded from the restricted quantitative synthesis but were considered for qualitative review where relevant. The primary outcome was progression to clinically diagnosed T1D during follow-up. Diagnosis was accepted according to recognised clinical, biochemical, or study-specific criteria reported by the original investigators. Secondary outcomes included progression to dysglycaemia, pre-symptomatic T1D stages, changes in glucose tolerance, and predictive performance measures such as hazard ratios, odds ratios, risk ratios, sensitivity, specificity, positive predictive value, negative predictive value, and discrimination indices. Eligible study designs included prospective cohort studies, longitudinal observational studies, nested case–control studies, and population-based screening studies reporting subsequent progression to T1D. Cross-sectional studies without longitudinal follow-up, case reports, case series, conference abstracts without sufficient data, editorials, narrative reviews, commentaries, and non-human studies were excluded. Studies not reporting outcomes related to T1D progression or insufficiently describing autoantibody exposure were also excluded. No geographical restrictions were applied during study selection. However, because one objective of this review was to evaluate the relevance of current evidence for African populations, studies from Africa and studies reporting findings relevant to African diabetes phenotypes were specifically identified, extracted, and synthesised separately where possible to facilitate evaluation of the applicability of existing prediction frameworks to African populations. Information sources A comprehensive and systematic literature search was conducted across four major electronic databases: PubMed, Scopus, Web of Science, and Embase. These databases were selected because of their extensive coverage of biomedical, endocrinology, epidemiology, and public health literature and their frequent use in systematic reviews of diagnostic and prognostic biomarkers. To maximise retrieval of relevant evidence, supplementary searches were undertaken through manual screening of reference lists of eligible studies, relevant review articles, and key publications identified during the screening process. The search covered studies published from January 2000 to March 2026. This timeframe was selected to capture contemporary evidence generated during the period of major advances in islet autoantibody identification, assay standardisation, disease staging frameworks, and biomarker-based prediction of type 1 diabetes [6, 10]. Searches were designed to identify studies evaluating the predictive value of islet autoantibodies, including insulin autoantibodies (IAA), islet cell antibodies (ICA), glutamic acid decarboxylase antibodies (GADA), insulinoma-associated antigen-2 antibodies (IA-2 A), zinc transporter 8 antibodies (ZnT8A), and related terminology used in earlier literature, including insulin-dependent diabetes mellitus (IDDM). No geographical restrictions were applied during the search process to ensure comprehensive identification of global evidence. However, studies conducted in African populations or reporting findings relevant to African diabetes phenotypes were specifically identified during data extraction and synthesis to support the review’s public health focus. Only studies published in English were included because of resource and feasibility constraints. The complete database-specific search strategies are provided in Supplementary File S1 to facilitate transparency, reproducibility, and independent verification of the search process. Search strategy A comprehensive search strategy was developed using a combination of controlled vocabulary terms (e.g., Medical Subject Headings [MeSH]) and free-text keywords related to type 1 diabetes, islet autoantibodies, disease prediction, and progression. The search strategy was designed to maximise sensitivity while maintaining relevance to the review objectives. Keywords and subject headings were identified through preliminary scoping searches, examination of key publications, and consultation of terminology commonly used in the type 1 diabetes prediction literature. Search terms included combinations of concepts relating to: (i) type 1 diabetes and related terminology (“type 1 diabetes”, “type 1 diabetes mellitus”, “insulin-dependent diabetes mellitus”, “IDDM”, “autoimmune diabetes”); (ii) islet autoantibodies (“islet autoantibodies”, “islet cell antibodies”, “ICA”, “insulin autoantibodies”, “IAA”, “glutamic acid decarboxylase antibodies”, “GADA”, “insulinoma-associated antigen-2 antibodies”, “IA-2A”, “zinc transporter 8 antibodies”, “ZnT8A”); and (iii) prediction-related outcomes (“prediction”, “risk”, “progression”, “incidence”, “screening”, “early detection”, “disease development”, “prognosis”). Boolean operators (AND, OR), truncation symbols, and database-specific indexing terms were used to optimise search sensitivity and specificity. Searches were adapted appropriately for each database according to indexing conventions and search functionalities. An example PubMed search strategy was: ((“Type 1 Diabetes Mellitus“[Mesh] OR “type 1 diabetes” OR “type 1 diabetes mellitus” OR “insulin-dependent diabetes mellitus” OR IDDM OR “autoimmune diabetes”)) AND. ((“Islet Cell Antibodies“[Mesh] OR “islet autoantibodies” OR “islet cell antibodies” OR ICA OR “insulin autoantibodies” OR IAA OR GADA OR “glutamic acid decarboxylase antibodies” OR IA-2 A OR “insulinoma-associated antigen-2 antibodies” OR ZnT8A OR “zinc transporter 8 antibodies”)) AND. ((“prediction” OR “risk” OR “progression” OR “incidence” OR “screening” OR “early detection” OR “disease development” OR prognosis)) The search strategy was iteratively refined through pilot searches to ensure adequate retrieval of known landmark studies, major longitudinal cohort investigations, and emerging studies from African populations, thereby supporting the review’s objective of evaluating the applicability of islet autoantibody-based prediction across diverse geographic settings. Complete database-specific search strategies are provided in Supplementary File S1 to facilitate transparency and reproducibility in accordance with PRISMA 2020 recommendations [27, 28]. Study selection All records retrieved from the database searches were imported into a reference management software package for organisation and duplicate removal. Following deduplication, the study selection process was conducted in two sequential stages: (i) title and abstract screening and (ii) full-text eligibility assessment. During the first stage, titles and abstracts were screened against the predefined eligibility criteria to identify studies potentially relevant to the review question. Studies were retained for full-text assessment if they investigated islet autoantibodies and their association with subsequent development of type 1 diabetes (T1D), disease progression, or risk prediction. Articles that were clearly unrelated to T1D, did not involve autoantibody assessment, were non-human studies, or represented reviews, editorials, conference abstracts without sufficient data, or case reports were excluded at this stage. In the second stage, the full texts of potentially eligible articles were retrieved and assessed against the inclusion and exclusion criteria described in Sect. 2.2. Reasons for exclusion at the full-text stage were documented and categorised to enhance transparency and reproducibility. A summary of excluded full-text studies and reasons for exclusion is provided in Supplementary Table S1. Studies meeting the eligibility criteria were included in the qualitative synthesis. Subsequently, studies reporting sufficiently comparable quantitative effect measures, such as risk ratios, hazard ratios, odds ratios, or cumulative incidence estimates, were assessed for eligibility for the restricted quantitative synthesis. The specific studies contributing to the quantitative synthesis are identified in Table 1 and Supplementary Table S2. Studies lacking comparable outcome metrics or reporting formats were retained in the qualitative review but excluded from the restricted quantitative synthesis. Table 1Characteristics and key findings of African studies included in the reviewStudyCountry/regionStudy type Sample sizeMain autoantibodiesKey findings relevant to predictionLutale et al. [18]TanzaniaCross-sectional245GADA, IA-2 AAutoantibody prevalence lower than reported in Caucasian populations; combined antibody testing improved identification of autoimmune diabetesAsanghanwa et al. [2]CameroonObservational302 diabetes casesGADA, IA-2 A, ZnT8AOnly 29% were autoantibody positive; antibody positivity predicted insulin requirement more strongly than clinical classificationBalcha et al. [3]EthiopiaCase-control236 T1D casesGADA, IA-2 A, ZnT8AMajority of young patients demonstrated immunogenetic characteristics consistent with autoimmune T1DPadoa et al. [26]South AfricaComparative cohort472GADA, IA-2 ALower IA-2 A prevalence and older age at diagnosis among Black South AfricansKatte et al. [14]Cameroon, Uganda, South AfricaMulticentre observational894GADA, IA-2 A, ZnT8AOnly 34.9% autoantibody positive; substantial evidence of non-autoimmune insulin-deficient diabetesKatte et al. [13]Sub-Saharan AfricaReviewMultiple studiesMultipleReported lower autoantibody prevalence (20–60%) and greater heterogeneity than in Western populationsSiraj et al. [33]AfricaEditorial/Public health reviewN/AN/AHighlighted unique diabetes challenges and evidence gaps across AfricaSvensson [34]Sub-Saharan AfricaCommentaryN/AN/AEmphasised emerging evidence of non-autoimmune insulin-deficient diabetes Characteristics and key findings of African studies included in the review Any uncertainties regarding study eligibility were resolved through discussion and consensus following re-examination of the full-text article and eligibility criteria. Particular attention was given to identifying longitudinal investigations evaluating progression from islet autoantibody positivity to clinically diagnosed T1D, as these provide the strongest evidence regarding predictive validity [22, 35]. In addition, studies conducted in African populations or reporting findings relevant to African diabetes phenotypes were specifically identified and coded during the extraction process to facilitate region-specific synthesis. The study selection process is summarised in the PRISMA 2020 flow diagram (Fig. 1), which details the numbers of records identified, screened, excluded, assessed for eligibility, and included in the qualitative synthesis and, where applicable, the restricted quantitative synthesis [28]. PRISMA 2020 flow diagram of study selection process Data extraction A standardised data extraction template was developed and piloted before formal extraction to ensure consistency and completeness across studies. Data extraction was undertaken systematically for all eligible studies included in the qualitative synthesis. The extraction framework was designed to support both narrative synthesis and restricted quantitative synthesis while facilitating detailed assessment of study characteristics, methodological quality, and applicability to African populations. For each study, the following information was extracted: author(s), publication year, country or geographic region, study design, recruitment setting, sample size, participant characteristics (including age group, baseline risk status, and family history where reported), duration of follow-up, and diagnostic criteria used for type 1 diabetes (T1D). Detailed information regarding islet autoantibody exposure was also collected, including the specific autoantibodies assessed, namely insulin autoantibodies (IAA), islet cell antibodies (ICA), glutamic acid decarboxylase antibodies (GADA), insulinoma-associated antigen-2 antibodies (IA-2 A), and zinc transporter 8 antibodies (ZnT8A), as well as the number of autoantibodies detected, antibody titres, affinity measures, persistence over time, and laboratory assay methods where available. Outcome data extracted included progression to clinically diagnosed T1D, progression to dysglycaemia or pre-symptomatic T1D stages, cumulative incidence estimates, and measures of predictive performance. Quantitative effect estimates, including hazard ratios (HRs), risk ratios (RRs), odds ratios (ORs), cumulative incidence rates, sensitivity, specificity, positive predictive values, negative predictive values, C-statistics, and other discrimination metrics, were extracted where reported. For studies contributing to the restricted quantitative synthesis, additional information required for effect size calculation, confidence interval estimation, and assessment of study comparability was collected. To enhance transparency, studies included in the qualitative synthesis were distinguished from those eligible for restricted quantitative synthesis. Only studies reporting sufficiently comparable outcome measures and effect estimates were included in the restricted quantitative synthesis. Studies lacking comparable quantitative data remained eligible for narrative synthesis and contextual interpretation. Information regarding the rationale for inclusion in the qualitative synthesis only, or inclusion in the restricted quantitative synthesis, was documented and reported in the supplementary materials. Studies contributing to the quantitative synthesis and the corresponding extracted effect measures are summarised separately in Supplementary Table S3. Given the specific public health focus of this review, contextual variables relevant to African populations were also extracted where available. These included country setting, population characteristics, autoantibody prevalence, diabetes phenotype, healthcare context, and any reported challenges related to diagnosis, classification, or screening implementation [13, 14, 33]. This enabled a more detailed evaluation of the applicability of existing evidence to African settings and facilitated region-specific interpretation of findings. A complete study-level extraction table summarising all included studies is provided in Supplementary Table S2 to enhance transparency and reproducibility of the evidence synthesis process. Outcomes of interest The primary outcome of interest was progression to clinically diagnosed type 1 diabetes (T1D) among individuals with detectable islet autoantibodies at baseline. T1D diagnosis was accepted according to the definitions and diagnostic criteria reported by the original studies, including clinical diagnosis, biochemical confirmation, requirement for insulin therapy, or established staging frameworks for T1D progression. The primary objective was to evaluate the predictive value of islet autoantibodies for identifying individuals at increased risk of developing T1D. Secondary outcomes included: (i) time to progression from islet autoantibody positivity to clinically diagnosed T1D; (ii) differences in risk according to the number, type, titre, affinity, and persistence of islet autoantibodies; and (iii) reported measures of predictive performance such as hazard ratios (HRs), risk ratios (RRs), odds ratios (ORs), cumulative incidence estimates, sensitivity, specificity, positive predictive value, negative predictive value, and discrimination indices (e.g., C-statistics). Particular attention was given to evidence demonstrating incremental risk associated with multiple-autoantibody positivity and high-affinity autoantibody responses, which have been identified as important predictors of disease progression [10, 22, 23]. Additional outcomes included progression to dysglycaemia, pre-symptomatic stages of T1D, and the performance of multivariable prediction models incorporating autoantibody profiles alongside clinical, demographic, or genetic factors [35]. These outcomes were included to provide a broader understanding of contemporary approaches to T1D risk prediction and disease staging. Given the public health focus of the review, subgroup outcomes relating to population characteristics, geographic region, and healthcare context were also examined. Particular emphasis was placed on evidence from African populations, which were extracted and synthesised separately where possible to evaluate the applicability of existing islet autoantibody-based prediction frameworks to African settings. These included studies reporting autoantibody prevalence, atypical diabetes phenotypes, preserved endogenous insulin secretion, non-autoimmune insulin-deficient diabetes, and challenges associated with implementation of autoantibody-based screening programmes [13, 14, 33, 34]. These outcomes were considered important for assessing the external validity and potential applicability of existing prediction models within African health systems. Collectively, these outcomes were selected to provide a comprehensive assessment of the predictive utility of islet autoantibodies, their clinical relevance for early identification of individuals at risk of T1D, and their broader implications for screening, prevention, and public health policy in Africa. Risk of bias assessment The methodological quality and risk of bias of included studies were assessed using structured tools appropriate for prognostic and observational research designs. For cohort studies, longitudinal observational studies, screening studies, and other non-randomised investigations, a domain-based risk-of-bias framework informed by ROBINS-I principles was used to evaluate potential sources of bias across key domains, including confounding, participant selection, classification of exposures, missing data, outcome measurement, and selective reporting. This approach is particularly relevant for studies evaluating prognostic biomarkers because differences in participant characteristics, follow-up procedures, and measurement methods may influence observed associations. For any interventional or randomised studies identified during the review process, the Cochrane Risk of Bias 2 (RoB 2) tool was applied to assess internal validity across domains, including randomisation procedures, deviations from intended interventions, missing outcome data, outcome measurement, and selective reporting. Although the majority of included studies were observational in nature, utilisation of study-design-specific tools ensured that methodological quality was evaluated appropriately across different evidence sources. Each study was assessed across the relevant domains and assigned an overall judgement of low, moderate, serious, or critical risk of bias in accordance with the recommendations of the respective assessment frameworks. Particular attention was given to potential confounding arising from family-history enrichment, participant selection procedures, variation in autoantibody assay methods, differential follow-up duration, and outcome ascertainment. These issues were considered especially important because many of the major type 1 diabetes prediction cohorts involve individuals at elevated baseline risk, such as first-degree relatives of affected individuals. To enhance transparency and reproducibility, domain-level assessments and overall risk-of-bias judgements were recorded for each included study. A complete study-by-study risk-of-bias assessment is provided in Supplementary Table S4, while an overall summary of risk-of-bias judgements is presented in Figure S2. Risk-of-bias findings were subsequently considered during evidence synthesis, interpretation of results, and assessment of certainty of evidence. Studies assessed as having a higher risk of bias were not automatically excluded but were considered during exploratory sensitivity assessments and qualitative interpretation of findings. The assessment process was conducted using predefined criteria to maximise consistency and align with recommended practices for systematic reviews and evidence synthesis [27, 28]. Effect measures A range of effect measures relevant to type 1 diabetes (T1D) prediction were extracted from eligible studies. These included hazard ratios (HRs), risk ratios (RRs), odds ratios (ORs), cumulative incidence estimates, and measures of predictive performance such as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and discrimination indices (e.g., C-statistics). The diversity of effect measures reflects the heterogeneous nature of the included literature, which encompassed prospective cohort studies, screening studies, biomarker validation studies, and prediction-model investigations. For studies reporting longitudinal progression to T1D, HRs were prioritised where available because they account for differences in follow-up duration and time-to-event information. Risk ratios and odds ratios were also extracted when reported by the original studies. Where appropriate, effect estimates were standardised for descriptive comparison and presentation. However, substantial heterogeneity in study design, outcome definitions, reporting formats, and analytical approaches limited the feasibility of formal quantitative synthesis across most studies. Studies eligible for restricted quantitative synthesis were required to report sufficiently comparable measures of risk associated with islet autoantibody positivity and subsequent progression to T1D. Consequently, only a restricted subset of studies reporting comparable risk estimates contributed to the quantitative synthesis. Studies reporting predictive accuracy metrics, discrimination indices, screening performance measures, or multivariable prediction models were synthesised narratively rather than quantitatively. This approach was adopted to minimise inappropriate pooling of fundamentally different outcome measures and to preserve methodological validity. For studies evaluating multiple autoantibody combinations, separate estimates were extracted for single-autoantibody positivity, multiple-autoantibody positivity, and high-risk autoantibody profiles where reported. Particular attention was given to evidence demonstrating incremental risk associated with increasing numbers of autoantibodies, higher autoantibody titres, and enhanced assay performance, consistent with previous research demonstrating that progression risk rises substantially as autoimmune burden increases [10, 22, 23]. To enhance transparency, studies contributing to the restricted quantitative synthesis were explicitly identified and distinguished from studies contributing exclusively to the qualitative synthesis. The studies contributing to the quantitative synthesis are summarised in Table 1 and Supplementary Table S3. This distinction was considered particularly important given the limited number of studies reporting directly comparable quantitative effect estimates suitable for inclusion in the restricted quantitative synthesis. Data synthesis Evidence synthesis was conducted using a sequential qualitative and quantitative approach. Given the substantial heterogeneity in study design, participant characteristics, autoantibody profiles, outcome definitions, follow-up duration, and reporting formats across the included literature, all eligible studies were first synthesised narratively. The qualitative synthesis focused on identifying consistent patterns regarding the predictive value of islet autoantibodies, progression to type 1 diabetes (T1D), risk stratification according to autoantibody profile, and implications for screening and early detection. Particular attention was given to studies conducted in African populations and studies reporting findings relevant to African diabetes phenotypes [13, 14, 33]. Restricted quantitative synthesis was undertaken only for studies reporting sufficiently comparable measures of association between islet autoantibody positivity and subsequent progression to T1D. Eligibility for inclusion in the restricted quantitative synthesis required comparable exposure definitions, outcome measures, and reporting formats. Because relatively few studies met these criteria, the quantitative synthesis was limited to a subset of studies providing directly comparable risk estimates. Consequently, the quantitative findings should be interpreted as complementary to, rather than a replacement for, the broader qualitative evidence base. Where quantitative synthesis was feasible, summary quantitative estimates were generated using a random-effects framework to account for anticipated between-study variability. Given the limited number of comparable studies and substantial clinical and methodological heterogeneity, these quantitative estimates were interpreted cautiously and considered supportive of, rather than definitive for, the broader qualitative evidence base. For studies reporting multiple eligible estimates, care was taken to avoid double-counting participants by selecting the most appropriate and independent comparison relevant to the review objective. Studies reporting predictive performance measures such as sensitivity, specificity, positive predictive value, negative predictive value, C-statistics, or multivariable prediction-model performance were synthesised narratively rather than quantitatively because of methodological differences that precluded direct pooling. Similarly, studies examining unique autoantibody subtypes, assay platforms, or region-specific disease phenotypes were incorporated into the narrative synthesis to provide contextual interpretation of findings. The synthesis also incorporated an Africa-focused evidence component. Findings from African studies were examined separately to identify differences in autoantibody prevalence, disease phenotype, progression patterns, and potential limitations of applying prediction models developed predominantly in European and North American populations. This approach was adopted in response to growing evidence that autoimmune and non-autoimmune forms of insulin-deficient diabetes may coexist in African settings and may influence the performance of conventional prediction strategies [13, 14, 34]. Overall, the combination of comprehensive narrative synthesis and restricted quantitative synthesis was considered the most appropriate strategy for preserving methodological validity while maximising the utility of available evidence. This approach is consistent with recommended practices for systematic reviews involving heterogeneous prognostic and biomarker studies [27, 28]. Heterogeneity assessment Statistical heterogeneity among studies included in the quantitative synthesis was assessed using Cochran’s Q statistic and the I² statistic. The Q statistic was used to evaluate whether observed variability in effect estimates exceeded that expected by chance alone, while the I² statistic quantified the proportion of total variation attributable to between-study heterogeneity. Conventional thresholds of approximately 25%, 50%, and 75% were used to indicate low, moderate, and high heterogeneity, respectively. A Q-test significance level of p < 0.10 was adopted because of the limited power of heterogeneity tests when only a small number of studies contribute to quantitative synthesis. Given the restricted number of studies eligible for quantitative synthesis, statistical measures of heterogeneity were interpreted cautiously. In such circumstances, I² and Q statistics may be unstable and should be considered complementary to clinical and methodological assessment rather than definitive indicators of inconsistency. Consequently, heterogeneity was evaluated not only quantitatively but also through detailed examination of study characteristics and sources of methodological variation. Potential sources of heterogeneity were explored by considering differences in study design, participant characteristics, baseline risk profile, family-history enrichment, geographic region, autoantibody type, number of autoantibodies assessed, laboratory assay methodology, outcome definitions, and duration of follow-up. Particular attention was given to variation between studies conducted in Europe and North America and the smaller number of studies conducted in African populations, where differences in disease phenotype, autoantibody prevalence, and classification of insulin-deficient diabetes have been increasingly reported [13, 14, 33, 34]. Where sufficient information was available, heterogeneity was further explored through subgroup and sensitivity analyses. However, because only a limited subset of studies contributed to the quantitative synthesis, these investigations were considered exploratory rather than confirmatory. The use of a random-effects framework was intended to account for anticipated between-study variability and provide more conservative summary estimates than would be obtained under a fixed-effect approach. Overall, interpretation of heterogeneity was informed by a combination of statistical measures, methodological appraisal, and clinical judgement to ensure that quantitative findings were evaluated within the broader context of the available evidence base. Subgroup analysis and exploration of heterogeneity To explore potential sources of heterogeneity and enhance the interpretation of findings, exploratory subgroup analyses were undertaken where sufficient information was available. Subgroup comparisons focused on clinically relevant characteristics identified a priori, including age group (children, adolescents, and adults), number of autoantibodies (single versus multiple-autoantibody positivity), type of autoantibody assessed (IAA, ICA, GADA, IA-2 A, and ZnT8A), geographic region (African versus non-African populations), and study setting or participant risk profile. These subgroup categories were selected based on previous evidence indicating that progression risk may vary according to autoantibody burden, demographic characteristics, and population context [10, 22, 23]. Particular attention was given to studies conducted in African populations because emerging evidence suggests important differences in diabetes phenotype, autoantibody prevalence, and patterns of insulin deficiency relative to populations from Europe and North America [13, 14, 33, 34]. Consequently, subgroup analyses were used primarily to explore potential contextual differences and inform the interpretation of the applicability of existing prediction models across settings. Given the limited number of studies contributing sufficiently comparable quantitative data, formal statistical investigation of study-level sources of heterogeneity was not considered robust. Instead, potential influences of study characteristics such as follow-up duration, sample size, participant risk profile, and assay methodology were explored descriptively through comparison of study findings and narrative synthesis. All subgroup analyses were interpreted as exploratory rather than confirmatory. The purpose of these analyses was not to establish definitive subgroup effects but rather to identify potential patterns that may warrant further investigation in future studies, particularly within underrepresented African populations. Findings from subgroup analyses were therefore interpreted alongside the broader qualitative evidence base rather than in isolation. This cautious approach was adopted to minimise overinterpretation of limited quantitative evidence while still providing clinically and policy-relevant insights regarding factors that may influence the predictive value of islet autoantibodies for type 1 diabetes development. Publication bias Potential publication bias was assessed using both graphical and statistical approaches. Funnel plots were generated to visually examine the distribution of study effect estimates against their corresponding standard errors. Visual inspection focused on identifying substantial asymmetry that might indicate small-study effects or selective publication of studies reporting stronger associations. In addition, exploratory statistical assessments were undertaken using Egger’s regression test and Begg’s rank correlation test. These methods were employed to provide supplementary evidence regarding the possible presence of publication bias or small-study effects. However, because only a limited number of studies contributed to the quantitative synthesis, the results of these tests were interpreted cautiously. Statistical methods for detecting publication bias generally have limited power when applied to quantitative syntheses involving only a small number of studies and may produce unstable or misleading results under such conditions. A trim-and-fill procedure was also considered as an exploratory method to evaluate the potential influence of unpublished or missing studies on summary quantitative estimates. Any adjusted estimates derived from this approach were interpreted descriptively rather than as definitive corrections because of the restricted number of studies contributing to the quantitative synthesis and the substantial methodological heterogeneity among included investigations. Consistent with recommendations for prognostic and biomarker reviews, publication bias assessments were regarded as supportive rather than confirmatory components of the evidence synthesis [27, 28]. Consequently, findings from funnel plots, Egger’s test, Begg’s test, and trim-and-fill procedures were interpreted alongside the broader characteristics of the evidence base, including study design, sample size, geographic representation, and methodological quality. Given the increasing volume of research on islet autoantibodies and type 1 diabetes prediction, and the possibility that studies reporting stronger predictive associations may be preferentially published, assessment of publication bias remained an important component of the review. Nevertheless, the results were interpreted conservatively and were not used as the sole basis for judging the credibility of the evidence. Certainty of evidence The overall certainty of evidence for each outcome was evaluated using the GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach, as developed by the GRADE Working Group. This framework assesses evidence across domains, including risk of bias, inconsistency, indirectness, imprecision, and publication bias [4, 36]. Evidence from observational studies was initially rated as low certainty but could be upgraded based on factors such as large effect size, dose–response relationship, and consistency across studies. Conversely, evidence could be downgraded due to methodological limitations or high heterogeneity. The GRADE approach was used to classify the certainty of evidence as high, moderate, low, or very low, thereby providing a transparent and structured basis for interpreting the strength of conclusions and informing public health recommendations. Study-level GRADE assessments and outcome-specific certainty ratings are presented in Supplementary Table S4 to enhance transparency and reproducibility. This is particularly important for translating findings on autoantibody-based prediction into policy-relevant screening strategies, especially in resource-constrained settings such as those found in many African countries. Results Study selection The systematic literature search across PubMed, Scopus, Web of Science, and Embase identified 1,842 records. An additional 37 records were identified through manual searches of reference lists of relevant reviews and eligible studies, yielding a total of 1,879 records. Following removal of duplicate records (n = 412), 1,467 unique records remained for title and abstract screening. During title and abstract screening, 1,238 records were excluded because they were not relevant to the review question. Common reasons for exclusion included absence of islet autoantibody assessment, non-human study design, lack of relevance to type 1 diabetes (T1D) prediction, absence of longitudinal follow-up, or failure to evaluate progression to T1D. The full texts of the remaining 229 articles were retrieved and assessed against the predefined eligibility criteria. Of these, 182 studies were excluded for the following reasons: cross-sectional design without longitudinal follow-up (n = 64), absence of relevant islet autoantibody exposure (n = 39), lack of outcome data relating to progression to T1D (n = 51), insufficient quantitative or methodological information for extraction (n = 18), and non-English publication (n = 10). A detailed summary of excluded studies and reasons for exclusion is provided in Supplementary Table S1. Ultimately, 47 studies met the eligibility criteria and were included in the qualitative synthesis. A complete list of included studies and extracted study characteristics is provided in Supplementary Table S2. Of the 47 included studies, only a small subset provided sufficiently comparable quantitative data, including directly extractable risk estimates or event data, and were therefore eligible for inclusion in the restricted quantitative synthesis. The remaining studies contributed to the qualitative synthesis because of substantial differences in study design, outcome definitions, reporting formats, predictive-performance metrics, or analytical approaches. The study selection process is summarised in the PRISMA 2020 flow diagram (Fig. 1). Adherence to PRISMA guidelines enhanced the transparency, reproducibility, and methodological rigour of the review process [27, 28]. The relatively high number of studies excluded at the full-text stage reflects the limited availability of longitudinal evidence examining progression from islet autoantibody positivity to clinically diagnosed T1D, particularly within African populations. This finding is consistent with previous reports highlighting the underrepresentation of African cohorts in T1D prediction research and the continuing need for region-specific longitudinal studies [13, 14, 33]. Figure 1 presents the process of study identification, screening, eligibility assessment, and inclusion. A total of 1,879 records were identified through database searching and supplementary sources. Following the removal of duplicate records, 1,467 studies underwent title and abstract screening, resulting in the exclusion of 1,238 records. Full-text assessment was conducted for 229 articles, of which 182 were excluded according to predefined eligibility criteria. Ultimately, 47 studies were included in the qualitative synthesis, while a restricted subset of studies provided sufficiently comparable quantitative data for inclusion in the restricted quantitative synthesis. The review process followed PRISMA 2020 recommendations to ensure transparency, reproducibility, and methodological rigour [27, 28]. Study characteristics The 47 studies included in the qualitative synthesis represented a diverse body of evidence evaluating the predictive value of islet autoantibodies for the development of type 1 diabetes (T1D) across different populations, age groups, and geographic settings. Most studies employed prospective cohort designs with longitudinal follow-up, providing the strongest evidence for assessing progression from autoantibody positivity to clinically diagnosed T1D. Several influential studies were conducted within large international screening and birth-cohort programmes, including TrialNet, TEDDY, DAISY, and related collaborative studies, which have substantially shaped current understanding of T1D prediction [22, 23, 35]. Other studies used nested case–control, observational, or screening-based designs, particularly among children and adolescents undergoing early-risk assessment [5, 32]. Collectively, the included studies spanned more than two decades of research, reflecting major advances in autoantibody detection technologies, disease staging frameworks, and predictive modelling approaches. Geographically, the evidence base was heavily concentrated in Europe and North America, with comparatively fewer studies originating from Africa, Asia, and other low- and middle-income regions. This distribution reveals an important geographic imbalance in the existing literature and raises concerns regarding the generalisability of prediction models developed predominantly in high-income populations. While the strongest longitudinal evidence originated from North American and European cohorts, several African studies identified important differences in disease presentation, autoantibody prevalence, and immunogenetic characteristics that may influence predictive performance and clinical interpretation. Table 1 summarises the African studies identified in this review. Findings from Tanzania, Cameroon, Ethiopia, South Africa, Uganda, and multicountry Sub-Saharan African cohorts consistently suggest greater heterogeneity in diabetes phenotype than is typically reported in European-origin populations [2, 3, 13, 14, 18, 26]. These studies collectively highlight the importance of region-specific validation of autoantibody-based prediction strategies. Study populations varied considerably with respect to age, baseline risk profile, and family history. Many of the largest prospective cohorts focused on children with increased genetic susceptibility or a family history of T1D, particularly first-degree relatives of affected individuals. Although these cohorts have generated invaluable evidence regarding disease progression, they may not fully represent the broader population in which most newly diagnosed T1D cases occur. As only a minority of individuals developing T1D have an affected first-degree relative, findings from family-enriched cohorts should be interpreted cautiously when extrapolating to population-wide screening programmes. Other studies included community-based populations, screening cohorts, and adult participants with latent autoimmune diabetes or slowly progressive autoimmune diabetes, thereby broadening the applicability of the evidence base [8, 31]. Across studies, the primary exposures of interest were islet autoantibodies, including insulin autoantibodies (IAA), islet cell antibodies (ICA), glutamic acid decarboxylase antibodies (GADA), insulinoma-associated antigen-2 antibodies (IA-2 A), and zinc transporter 8 antibodies (ZnT8A). Technological advances in assay methodology, particularly electrochemiluminescence-based platforms, improved sensitivity and specificity and enabled more refined assessment of antibody affinity and titre [6, 10]. Several studies demonstrated that predictive accuracy improved substantially when multiple autoantibodies were considered simultaneously rather than individually, supporting contemporary staging frameworks for pre-symptomatic T1D [11, 22, 23]. Outcome measures were generally centred on progression to clinically diagnosed T1D, although definitions and reporting approaches varied across studies. Some investigations evaluated progression to dysglycaemia or intermediate stages of disease, while others reported diagnostic performance measures such as sensitivity, specificity, positive predictive value, and discrimination statistics. A limited subset of studies provided sufficiently comparable risk estimates or event-level data for inclusion in the quantitative synthesis. The majority contributed to the qualitative synthesis because of heterogeneity in study design, outcome definitions, follow-up duration, and reporting formats. This distinction is important when interpreting pooled estimates and reflects the broader methodological diversity of the evidence base. Overall, the included studies demonstrate substantial heterogeneity in design, population characteristics, geographic distribution, and analytical approaches. Nevertheless, a consistent pattern emerged across both high-income and African settings: the presence of multiple islet autoantibodies and higher antibody titres was associated with increased risk of progression to T1D. At the same time, evidence from African populations revealed important variations in autoantibody prevalence, age at diagnosis, endogenous insulin secretion, and disease phenotype that may influence the performance of prediction models developed elsewhere. These findings underscore the need for greater representation of African populations in future longitudinal studies and provide important context for interpreting the subsequent qualitative and quantitative syntheses. Risk of bias results Risk-of-bias assessments were completed for all 47 studies included in the qualitative synthesis using the methodological framework described in Sect. 2.8. Detailed study-level assessments are presented in Supplementary Table S3, while an overall summary of domain-specific judgements is shown in Figure S2. Overall, methodological quality was generally moderate to high, reflecting the predominance of prospective cohort studies with longitudinal follow-up. Most studies were classified as having low or moderate risk of bias, whereas only a limited number were judged to have serious risk of bias in one or more domains. Across the confounding domain, the majority of large prospective cohorts adequately accounted for important determinants of T1D progression, including age, sex, genetic susceptibility, family history, and baseline metabolic characteristics [22, 23, 35]. However, several studies, particularly those conducted in resource-constrained settings, reported limited adjustment for environmental, socioeconomic, and healthcare-access factors that may influence both autoantibody expression and progression risk [13, 33]. Consequently, residual confounding cannot be entirely excluded, especially when comparing findings across diverse geographic regions. Selection bias was generally low to moderate. Most longitudinal studies recruited participants using clearly defined eligibility criteria and established follow-up procedures. Nevertheless, several influential cohorts, including major screening and prediction programmes, enrolled individuals with increased genetic susceptibility or a family history of T1D. While these cohorts provide valuable insight into disease progression, they may not fully represent the broader population because most individuals who develop T1D do not have an affected first-degree relative. This consideration is particularly important when interpreting the generalisability of risk estimates to population-based screening programmes. Risk of bias arising from exposure measurement was generally low. Most studies used validated laboratory methods to assess islet autoantibodies, including IAA, ICA, GADA, IA-2 A, and ZnT8A. Recent studies employing electrochemiluminescence-based assays demonstrated particularly high measurement reliability and improved discrimination of high-risk individuals [6, 10]. Nevertheless, variation in assay platforms, positivity thresholds, and laboratory protocols contributed to methodological heterogeneity across studies and may partially explain differences in reported effect estimates. Outcome measurement bias was also judged to be low in most studies because progression to T1D was generally defined using recognised clinical and biochemical diagnostic criteria. Longitudinal follow-up allowed for prospective ascertainment of incident disease and reduced the likelihood of reverse causation. However, some studies had relatively short follow-up periods, which may have underestimated progression among individuals with slower disease trajectories or lower-risk autoantibody profiles [19]. Missing data and attrition bias were generally low in large cohort studies, many of which reported structured follow-up procedures and retention rates. Nevertheless, a subset of studies reported incomplete follow-up or provided limited information regarding the handling of missing data. Because progression to T1D may occur over extended periods, attrition remains a potentially important source of bias in long-term observational studies. Reporting bias was judged to be low to moderate across the evidence base. Most studies provided sufficient methodological detail to permit the extraction of relevant outcomes and assessment of study quality. However, relatively few studies reported pre-specified analytical protocols, and selective reporting of statistically significant findings cannot be entirely excluded, particularly among smaller studies. In addition, some studies reported predictive-performance measures without providing effect estimates suitable for quantitative synthesis, thereby limiting comparability across investigations. Among the African studies included in the review, methodological quality was generally moderate. The principal limitations involved smaller sample sizes, shorter follow-up periods, and limited availability of longitudinal outcome data, rather than deficiencies in laboratory assessment of autoantibodies. These constraints reflect broader challenges associated with conducting long-term diabetes prediction research in resource-limited settings and highlight the need for additional prospective African cohorts. Overall, the risk-of-bias assessment indicates that the evidence base provides a reasonably reliable foundation for qualitative synthesis and supports cautious interpretation of the restricted quantitative findings. However, the presence of moderate heterogeneity, family-history enrichment in several cohorts, and the limited number of longitudinal African studies underscore the importance of interpreting pooled estimates within the broader context of the available evidence rather than as definitive measures of population-wide risk. Qualitative synthesis The qualitative synthesis of the 47 included studies revealed a consistent and clinically important pattern: islet autoantibodies are among the strongest currently available predictors of progression to type 1 diabetes (T1D). Across prospective cohorts, screening programmes, longitudinal observational studies, and biomarker-validation investigations, the presence of one or more islet autoantibodies consistently preceded the development of clinically diagnosed T1D by months or years. This relationship was particularly evident in paediatric populations, where longitudinal follow-up enabled detailed characterisation of disease progression from normoglycaemia through dysglycaemia to overt T1D [19, 22, 23]. Similar predictive relationships were also observed among adults with autoimmune forms of diabetes, including latent autoimmune diabetes in adults and slowly progressive autoimmune diabetes, although progression rates were generally more heterogeneous than those observed in childhood-onset disease [8, 31]. A central finding across the literature is that predictive performance depends not merely on the presence of autoantibodies but on their number, type, titre, affinity, and persistence. Individuals with a single autoantibody generally exhibited lower and more variable progression risk, whereas those with multiple autoantibodies demonstrated substantially higher probabilities of developing T1D. This finding was remarkably consistent across large prospective cohorts and supports contemporary staging frameworks in which multiple-autoantibody positivity represents established pre-symptomatic autoimmune disease [11, 15, 16]. Furthermore, several studies demonstrated that quantitative assessment of antibody titres and affinity improved risk discrimination beyond simple positive-versus-negative classification. High-affinity ZnT8A responses, elevated antibody titres, and assay-enhanced detection strategies consistently identified individuals at particularly high risk of progression [6, 10, 22, 23]. Collectively, these findings suggest that islet autoantibodies are most informative when interpreted as part of a multidimensional risk framework rather than as isolated laboratory markers. The literature also highlights the growing clinical importance of early detection and screening. Several studies demonstrated that identification of autoantibody-positive individuals enables closer monitoring, earlier diagnosis, and participation in emerging prevention or disease-delay programmes [5, 17]. Screening initiatives conducted in community and population settings further suggest that risk identification can be implemented beyond specialised research centres, although important challenges remain regarding infrastructure, follow-up pathways, ethical considerations, and resource allocation [32]. These findings reinforce the translational potential of autoantibody-based screening and support ongoing efforts to incorporate biomarker-guided approaches into broader public health strategies. A particularly important contribution of the present review is the synthesis of emerging evidence from African populations. Unlike many previous reviews that primarily relied on European and North American data, this review identified a growing body of African evidence demonstrating that the epidemiology and phenotype of T1D may differ substantially across settings. Studies from Tanzania, Cameroon, Ethiopia, South Africa, Uganda, and multicountry Sub-Saharan African cohorts reported considerable heterogeneity in autoantibody prevalence and disease presentation [2, 3, 18, 26]. In several African cohorts, autoantibody positivity was observed less frequently than would typically be expected in classical autoimmune T1D populations from Europe and North America. At the same time, many patients displayed clinical features consistent with insulin deficiency despite limited evidence of autoimmune destruction. These findings suggest that the spectrum of insulin-deficient diabetes in Africa may be broader than conventional autoimmune models imply. The strongest evidence supporting this interpretation emerged from recent investigations conducted within the Young-Onset Diabetes in Sub-Saharan Africa (YODA) programme. Katte et al. [14] reported substantial evidence of non-autoimmune insulin-deficient diabetes among children and young adults, while Katte et al. [13], Siraj et al. [33], and Svensson [34] highlighted the growing recognition of atypical diabetes phenotypes across the region. These observations have important implications for predictive modelling because risk algorithms developed predominantly in high-income countries may not perform identically in African populations. Consequently, although islet autoantibodies remain valuable biomarkers in African settings, their interpretation may require contextual calibration and integration with local clinical, metabolic, and epidemiological characteristics. Another recurring theme across the literature was the increasing movement toward integrated prediction models. Rather than relying solely on autoantibody positivity, contemporary studies increasingly combine immunological markers with genetic susceptibility, demographic characteristics, metabolic indicators, and clinical risk factors to improve predictive performance [7, 35]. These approaches appear particularly promising for addressing heterogeneity in disease progression and may ultimately improve identification of individuals most likely to benefit from surveillance or preventive interventions. Nevertheless, important evidence gaps remain regarding the prediction of progression rate, identification of atypical phenotypes, and validation of prediction tools in underrepresented populations. The qualitative synthesis also revealed a notable imbalance in the geographic distribution of evidence. Most longitudinal studies were conducted in Europe and North America, whereas comparatively few originated from Africa despite the region’s growing diabetes burden [9, 24, 25]. This imbalance limits understanding of how predictive biomarkers perform across diverse populations and reinforce the need for additional African cohort studies with long-term follow-up. The relatively small number of African investigations identified in the present review should therefore be interpreted as evidence of a significant research gap rather than evidence against the clinical utility of islet autoantibodies within African populations. Overall, the qualitative evidence consistently supports the role of islet autoantibodies as important predictors of progression to T1D. The literature demonstrates strong biological plausibility, consistent longitudinal associations, and increasing utility within risk-stratification frameworks. However, the review also highlights important differences in disease phenotype, autoantibody prevalence, and healthcare context across regions, particularly within Africa. These findings support the continued development of autoantibody-based prediction strategies while emphasising the need for region-specific validation and greater inclusion of African populations in future research. The qualitative synthesis, therefore, provides a critical foundation for interpreting the subsequent quantitative findings and for understanding their implications for clinical practice and public health policy. Quantitative synthesis findings The quantitative synthesis included a limited subset of studies that reported sufficiently comparable measures of association between islet autoantibody characteristics and progression to type 1 diabetes (T1D). Consequently, the findings presented in Table 2 should be interpreted as exploratory quantitative evidence that complements the broader qualitative synthesis rather than as definitive pooled estimates representative of all included studies. Table 2Summary of studies contributing to the quantitative synthesisStudyPopulation/designAutoantibody exposure/comparisonFollow-upExtracted quantitative dataDerived effect used in synthesisNg et al. [22, 23]1,481 autoantibody-positive children from harmonised prospective birth-risk cohortsSingle autoantibody-positive children with antibody titre at or above type-specific threshold vs. below threshold5 yearsn = 364 vs. 590; 5-year diabetes risk = 21.9% vs. 6.1%Exploratory RR = 3.60 (95% CI 2.49–5.22)Ng et al. [22, 23]527 multiple-autoantibody-positive children from harmonised prospective cohortsThree titres above threshold vs. zero titres above threshold5 yearsn = 60 vs. 49; 5-year diabetes risk = 75.1% vs. 24.7%Exploratory RR = 3.06 (95% CI 1.83–5.11)Jia et al. [10]DAISY cohort; children with persistent single ZnT8A positivityECL-confirmed high-affinity ZnT8A vs. ECL-negative single ZnT8AMedian 7.2 years for ECL-negative group; progression reported during follow-up3/3 progressed in ECL-confirmed group vs. 0/8 in ECL-negative groupExploratory RR = 15.75 (95% CI 1.04–238.88)*Jia et al. [10]ASK + DAISY / screening-validation studyECL-confirmed ZnT8A vs. RBA-only ZnT8A2 years2-year cumulative risk = 24% vs. 2%; overall PPV of ECL-ZnT8A = 87.1% vs. 53.5% for RBA-ZnT8AQuantitative summary only; not pooled with above due different denominator structureNg et al. [23]1,403 autoantibody-positive children; 523 developed diabetesModel using autoantibody levels vs. positivity-only indicators10 years (main model)C-index = 0.76 (95% CI 0.74–0.77) for levels vs. 0.72 (95% CI 0.71–0.74) for positivity-only; 2-year C-index = 0.82 (95% CI 0.81–0.83)Predictive-discrimination summary only; not pooled with RR studiesTempleman et al. [35]7,798 TEDDY and 4,068 TrialNet participantsModel using genetic risk score, age, family history, and autoantibody statusVariable cohort follow-up305/7,798 developed T1D in TEDDY; 1,373/4,068 in TrialNet; recalibration improved Brier score from 0.40 to 0.16 in TrialNetPrediction-model summary only; not pooled with RR studies Summary of studies contributing to the quantitative synthesis Across the studies contributing quantitative data, a consistent pattern was observed whereby higher autoantibody titres, multiple-autoantibody positivity, and enhanced assay-based detection methods were associated with increased risk of progression to T1D. In the harmonised prospective cohort analyses reported by Ng et al. [22], children with single-autoantibody positivity and antibody titres above predefined thresholds exhibited substantially greater 5-year progression risk than those with titres below threshold (21.9% versus 6.1%), corresponding to an exploratory risk ratio of 3.60 (95% CI: 2.49–5.22). Similarly, among children with multiple-autoantibody positivity, the presence of three titres above threshold was associated with a 5-year diabetes risk of 75.1% compared with 24.7% among those without elevated titres, yielding an exploratory risk ratio of 3.06 (95% CI: 1.83–5.11). These findings support the concept that quantitative characteristics of autoantibody responses contribute important prognostic information beyond simple positivity status. Additional evidence regarding the importance of assay characteristics was provided by Jia et al. [10]. In the DAISY cohort, all participants with electrochemiluminescence-confirmed high-affinity ZnT8A progressed to T1D during follow-up, whereas no progression was observed among individuals with ECL-negative ZnT8A. Although the resulting exploratory risk estimate (RR = 15.75; 95% CI: 1.04–238.88) was associated with substantial uncertainty because of the small sample size, the direction of effect suggests that high-affinity autoantibody detection may identify individuals with particularly elevated progression risk. Consistent findings were observed in the associated screening-validation analysis, where ECL-confirmed ZnT8A demonstrated higher predictive performance than conventional radiobinding assay detection. Several studies also highlighted the value of integrating quantitative autoantibody information into broader prediction frameworks. Ng et al. [23] reported improved predictive discrimination when quantitative autoantibody levels were incorporated into risk models compared with positivity-only approaches, increasing the C-index from 0.72 to 0.76 over 10 years. Similarly, Templeman et al. [35] demonstrated that prediction models combining autoantibody status with genetic, demographic, and clinical information achieved improved calibration and discrimination across large prospective cohorts. These findings suggest that autoantibody profiles may contribute most effectively when incorporated into multifactorial prediction strategies rather than used in isolation. Interpretation of the quantitative findings should nevertheless be undertaken cautiously. Only a limited number of studies reported directly comparable effect measures suitable for quantitative synthesis, and substantial methodological differences existed with respect to population characteristics, follow-up duration, assay methodology, and outcome definitions. Furthermore, most quantitative evidence originated from Europe and North America, whereas relatively little comparable longitudinal evidence was available from African populations. Consequently, the quantitative synthesis should be viewed primarily as supportive evidence reinforcing the broader patterns identified in the qualitative review. From an African public health perspective, the findings are particularly important because they demonstrate the potential utility of islet autoantibody-based prediction while simultaneously highlighting important evidence gaps. Emerging African studies suggest that insulin-deficient diabetes may not always follow classical autoimmune pathways and that autoantibody prevalence may differ from patterns observed in high-income settings [13, 14, 34]. As a result, the applicability of existing prediction algorithms to African populations remains uncertain and requires further validation through prospective regional cohort studies. Overall, the quantitative synthesis supports the broader conclusion that islet autoantibodies are important predictors of progression to T1D and that predictive performance improves when multiple antibodies, quantitative titre information, or enhanced assay methodologies are considered. However, the limited number of directly comparable studies and the underrepresentation of African populations indicate that these findings should be interpreted as complementary to the qualitative evidence rather than as definitive estimates of population-level risk. Figure 2 presents study-specific risk estimates from the limited number of studies reporting sufficiently comparable quantitative outcomes. Across studies, higher autoantibody titres, multiple-autoantibody positivity, and high-affinity ZnT8A were consistently associated with increased risk of progression to type 1 diabetes. The wide confidence interval observed for the ZnT8A analysis reflects the small sample size and resulting statistical uncertainty of that study. Given the restricted number of comparable studies and methodological heterogeneity, the figure should be interpreted as supportive quantitative evidence that complements the broader qualitative synthesis rather than as a definitive estimate of population-level risk. Forest plot of studies included in the restricted quantitative synthesis Exploratory subgroup and sensitivity assessments Given the limited number of studies contributing directly comparable quantitative data, subgroup and sensitivity analyses were interpreted as exploratory assessments intended to provide contextual insight rather than definitive statistical comparisons. These assessments were undertaken to examine whether the observed associations between islet autoantibodies and progression to type 1 diabetes (T1D) were broadly consistent across different autoantibody profiles, population groups, geographic settings, and methodological characteristics. Across the included studies, a consistent pattern emerged whereby progression risk increased with the number of detectable autoantibodies. Evidence from the harmonised prospective cohorts reported by Ng et al. [22, 23] demonstrated that children with multiple elevated autoantibody titres experienced substantially higher progression rates than those with single-autoantibody positivity. These observations are consistent with contemporary staging frameworks for T1D, which recognise multiple-autoantibody positivity as a marker of established pre-symptomatic autoimmune disease [11, 15, 16]. Although the limited number of quantitatively comparable studies precluded formal subgroup pooling, the direction of findings was consistent across investigations. Variation was also observed according to autoantibody type and assay methodology. Studies evaluating ZnT8A, particularly using electrochemiluminescence-based assays, generally reported stronger predictive performance than studies relying solely on conventional autoantibody detection methods [6, 10]. Similarly, studies incorporating combinations of GADA, IA-2 A, ZnT8A, and IAA consistently reported improved risk discrimination compared with single-marker approaches. These findings suggest that both autoantibody profile and assay sensitivity may influence predictive performance, although the available quantitative evidence remains insufficient for definitive comparative conclusions. Differences were also observed across age groups. Studies involving children generally reported a more predictable progression from autoantibody positivity to clinically diagnosed T1D than studies involving adults. This observation is consistent with the classical autoimmune disease trajectory typically observed in paediatric populations [12, 30]. By contrast, adult cohorts often included individuals with latent autoimmune diabetes in adults or slowly progressive autoimmune diabetes, resulting in greater variability in progression rates and clinical presentation [8, 31]. Consequently, age-related differences may contribute to some of the heterogeneity observed across studies. Particularly noteworthy were the differences observed between African and non-African studies. The strongest evidence for autoantibody-based prediction originated from European and North American cohorts characterised by relatively homogeneous autoimmune phenotypes and established screening infrastructures. In contrast, studies from Sub-Saharan Africa reported greater variability in autoantibody prevalence, endogenous insulin secretion, and clinical presentation [13, 14, 33, 34]. These observations suggest that predictive models developed in high-income settings may require contextual adaptation before implementation in African populations. Because only a limited number of African studies were identified, formal quantitative comparison between regions was not feasible; nevertheless, the qualitative evidence consistently highlighted the importance of regional calibration and validation. Additional sensitivity assessments focused on methodological characteristics of the included studies. Comparisons across studies suggested that the overall direction of association between autoantibody positivity and T1D progression remained consistent irrespective of study size, follow-up duration, or overall risk-of-bias classification. Large prospective cohorts generally reported findings similar to those observed in smaller studies, although longer follow-up periods naturally captured higher cumulative progression rates. Likewise, exclusion of studies reporting non-comparable predictive-performance measures did not materially alter the overall interpretation of the evidence base. Taken together, these exploratory assessments indicate that the observed relationship between islet autoantibodies and progression to T1D is broadly consistent across multiple study contexts. However, the restricted number of studies contributing directly comparable quantitative outcomes, together with substantial methodological and geographic heterogeneity, means that these findings should be interpreted cautiously. Rather than providing definitive subgroup-specific estimates, the analyses primarily highlight areas where future research, particularly within African populations, is needed to strengthen understanding of context-specific prediction and risk stratification. Publication bias findings Publication bias was assessed using visual inspection of the funnel plot (Figure S3) together with exploratory consideration of potential small-study effects. However, because only a limited number of studies contributed directly comparable quantitative data, formal statistical assessment of publication bias should be interpreted cautiously. Current methodological guidance indicates that funnel-plot asymmetry and associated statistical tests have limited reliability when applied to small numbers of studies and heterogeneous evidence bases [27, 28]. Visual inspection of the funnel plot suggested some degree of asymmetry, with smaller studies tending to report larger effect estimates than larger cohort investigations. This pattern is commonly observed in biomarker-prediction research and may reflect several factors, including publication bias, selective reporting, methodological differences, or the enrichment of high-risk individuals within smaller specialised cohorts. Nevertheless, the small number of quantitatively comparable studies makes it difficult to distinguish confidently between these explanations. Formal statistical approaches, including Egger’s regression and Begg’s rank correlation procedures, were considered exploratory because of the restricted evidence base. Consequently, the results of these tests were not used as primary evidence either for or against the presence of publication bias. Similarly, trim-and-fill procedures were not considered sufficiently reliable to support definitive inference given the limited number of studies available for quantitative synthesis. Several additional considerations support cautious interpretation of potential publication bias. First, a substantial proportion of the literature on islet autoantibodies reports diagnostic-performance measures, prediction-model metrics, or other outcomes that cannot readily be combined into conventional meta-analytic effect estimates. As a result, selective availability of quantitative data may contribute to synthesis-level bias even when study-level publication bias is minimal. Second, studies reporting stronger predictive associations may be more likely to be published, cited, or selected for detailed reporting than studies reporting weaker or null associations. Third, important geographic disparities remain evident within the literature, with relatively few studies originating from African populations despite the growing burden of diabetes across the region [13, 14, 33]. Taken together, the publication-bias assessment does not provide strong evidence that the observed findings are primarily driven by selective reporting. However, neither does it allow publication bias to be confidently excluded. Consequently, the results should be interpreted within the broader context of the qualitative evidence base, the limited number of quantitatively comparable studies, and the continuing underrepresentation of African populations in longitudinal diabetes prediction research. Continued efforts to improve reporting transparency, protocol registration, and inclusion of diverse populations will strengthen future evidence syntheses in this field. Discussion This systematic review synthesised evidence on the predictive value of islet autoantibodies for progression to type 1 diabetes (T1D), with particular attention to the implications of this evidence for African populations. Across the included studies, a consistent pattern emerged whereby the presence of islet autoantibodies, especially multiple-autoantibody positivity and elevated antibody titres, was associated with increased risk of progression to clinically diagnosed T1D. The qualitative evidence was remarkably consistent across diverse study settings and populations, while the restricted quantitative synthesis provided supportive evidence that higher autoantibody burden and enhanced assay-based detection methods are associated with greater progression risk. Collectively, the findings reinforce the central role of islet autoantibodies as important biomarkers of autoimmune disease progression and support their utility within contemporary risk-stratification frameworks. However, the review also highlights important differences in study populations, assay methodologies, and geographic settings that should be considered when interpreting predictive performance. The findings are broadly consistent with previous literature demonstrating that islet autoantibodies are among the earliest detectable indicators of autoimmune beta-cell destruction and subsequent T1D development [11, 15]. The observation that progression risk increases with the number of autoantibodies aligns closely with current staging models of pre-symptomatic T1D and with evidence from major longitudinal cohorts demonstrating that multiple-autoantibody positivity is associated with substantially greater disease risk than single-autoantibody positivity [16, 19]. Similarly, the improved predictive performance associated with quantitative antibody assessment, antibody affinity, and enhanced detection technologies is consistent with emerging evidence that autoantibody characteristics provide prognostic information beyond simple positive-versus-negative classification [6, 10, 22, 23]. The findings also support ongoing efforts to integrate immunological markers with genetic, metabolic, and clinical risk factors within multifactorial prediction models [35]. An important observation emerging from the review is that much of the strongest longitudinal evidence originates from cohorts enriched for genetic susceptibility or family history of T1D, including large studies such as TrialNet, TEDDY, and DAISY. These cohorts have generated critical insights into disease progression and biomarker performance, but they may not fully represent the broader population because most newly diagnosed individuals with T1D do not have an affected first-degree relative. Consequently, risk estimates derived from family-enriched cohorts may not be directly transferable to population-wide screening programmes. Future research should therefore continue to evaluate autoantibody performance in more representative community-based populations, particularly in regions where family-based screening infrastructure is limited. The African evidence identified in this review provides one of its most important contributions. Although relatively few African studies met the inclusion criteria, the available evidence consistently suggests that the epidemiology and phenotype of insulin-deficient diabetes in Sub-Saharan Africa may differ from patterns commonly reported in Europe and North America. Studies from Tanzania, Cameroon, Ethiopia, South Africa, Uganda, and multicountry African cohorts indicate substantial heterogeneity in autoantibody prevalence, age at diagnosis, and clinical presentation [2, 3, 13, 14, 18, 26]. In particular, recent findings from the YODA programme suggest that a considerable proportion of insulin-deficient diabetes among African children and young adults may occur in the absence of detectable islet autoimmunity [14]. These observations do not diminish the value of islet autoantibodies as predictive biomarkers; rather, they indicate that prediction models developed in predominantly European-origin populations may require adaptation and validation before implementation in African settings. The public health implications of these findings vary across population groups and healthcare systems. In high-income settings with established screening infrastructure, autoantibody-based risk assessment may facilitate earlier diagnosis, reduce the likelihood of diabetic ketoacidosis at presentation, and improve identification of individuals eligible for preventive or disease-delay interventions [5, 17]. In African settings, the potential benefits of early detection are similarly important but must be balanced against practical considerations relating to laboratory infrastructure, workforce capacity, affordability, and continuity of follow-up care. Furthermore, because diabetes phenotypes appear more heterogeneous in parts of Sub-Saharan Africa, implementation of screening programmes should be accompanied by locally relevant validation studies and ongoing surveillance of clinical outcomes. These considerations highlight the importance of integrating biomarker research with broader health-system strengthening initiatives rather than viewing screening as a stand-alone intervention. The present review possesses several important strengths. It followed PRISMA 2020 recommendations, employed a comprehensive multi-database search strategy, incorporated structured risk-of-bias assessment procedures, and synthesised evidence from a broad range of study designs and populations [27, 28]. In addition, the review extends previous syntheses by explicitly examining the implications of autoantibody-based prediction for African populations, a topic that remains underrepresented in the existing literature. Nevertheless, several limitations warrant consideration. First, only a limited number of studies reported sufficiently comparable quantitative outcomes to permit restricted quantitative synthesis, and therefore, the quantitative findings should be interpreted as supportive rather than definitive evidence. Second, substantial heterogeneity existed across studies with respect to populations, assay methodologies, follow-up duration, outcome definitions, and reporting approaches. Third, much of the available evidence originated from Europe and North America, whereas relatively few longitudinal studies were conducted in Africa. This geographic imbalance limits certainty regarding the transferability of existing prediction models to African populations and represents an important evidence gap. Fourth, although publication-bias assessments did not reveal strong evidence of systematic reporting bias, the limited number of quantitatively comparable studies restricts the reliability of such assessments. Finally, the review relied on published evidence and is therefore inherently dependent on the quality and completeness of available studies. Overall, the findings indicate that islet autoantibodies remain important predictors of progression to T1D and have considerable potential for informing risk stratification and early detection strategies. At the same time, the review highlights the need for greater representation of African populations in longitudinal prediction research and for careful validation of existing prediction frameworks across diverse settings. Future studies should prioritise prospective African cohorts, standardised autoantibody assessment protocols, and integrated prediction models that account for regional variation in diabetes phenotype. Such efforts will strengthen the evidence base needed to support equitable and contextually appropriate approaches to T1D prediction and prevention. Conclusion This systematic review synthesised current evidence regarding the predictive value of islet autoantibodies for progression to type 1 diabetes (T1D) and examined the implications of this evidence for African populations. Across the included studies, a consistent pattern emerged whereby multiple-autoantibody positivity, higher autoantibody titres, and enhanced assay-based detection methods were associated with increased risk of progression to clinically diagnosed T1D. The findings further suggest that quantitative assessment of autoantibody characteristics may provide additional prognostic information beyond simple positive-versus-negative classification and may contribute to more refined approaches to risk stratification. An important contribution of this review is the identification of emerging evidence from Sub-Saharan Africa, where diabetes phenotypes appear more heterogeneous than those commonly reported in high-income settings. Although islet autoantibodies remain relevant biomarkers within African populations, the available evidence indicates that prediction models developed predominantly in Europe and North America may require contextual validation and adaptation before widespread implementation. These observations reinforce the importance of considering regional epidemiology, healthcare infrastructure, and disease heterogeneity when translating biomarker-based prediction into public health practice. Several limitations should be acknowledged when interpreting the findings. Only a limited number of studies reported sufficiently comparable quantitative outcomes to permit restricted quantitative synthesis, and substantial heterogeneity existed across study populations, assay methodologies, follow-up durations, and reporting approaches. Furthermore, African populations remain underrepresented within the longitudinal prediction literature, limiting certainty regarding the performance of existing prediction frameworks across diverse settings. These limitations highlight important evidence gaps rather than weaknesses in the underlying biological rationale linking islet autoimmunity to T1D progression. Overall, the available evidence supports the continued investigation and application of islet autoantibodies as important biomarkers of T1D risk. However, further prospective studies, particularly in African populations and other underrepresented settings, are needed to strengthen the evidence base, improve external validity, and refine context-specific prediction strategies. Future research should prioritise longitudinal African cohorts, standardised autoantibody assessment protocols, and integrated prediction models that combine immunological, genetic, metabolic, and clinical information. Such efforts will contribute to a more comprehensive understanding of T1D progression and support the development of equitable, evidence-informed approaches to early detection and prevention. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Author contributions B.I.G.A is the corresponding author who drafted the manuscript, systematic review search and contributed to all the sections. A. A. A, F. L. D, M. O. A, F. O. O, G. A. F, M. C. A. O, N. A. O, O. D. A, E. O. O, U. I. D, N. N. U, E. U. O, R. F. A, K. E. I and L. W. A contributed to the systematic review search, Discussion and entire manuscript review. Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Data availability The data supporting the findings of this study are derived from previously published articles and publicly available databases. All relevant data are included within the manuscript and its supplementary materials. Additional information, including extracted datasets and analytic details, is available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate This study is a systematic review and meta-analysis based exclusively on previously published studies and publicly available data. As such, it did not involve primary data collection or human participants and therefore did not require ethical approval or informed consent. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement The data supporting the findings of this study are derived from previously published articles and publicly available databases. All relevant data are included within the manuscript and its supplementary materials. Additional information, including extracted datasets and analytic details, is available from the corresponding author upon reasonable request.